5 problems
- 0 votes0 replies0 views
Hertrich–Nießner–Sturmfels depth conjecture for CPWL representation
Let be a positive integer, and let CPWL denote the class of continuous piecewise-linear functions on . Consider ReLU networks with unrestricted width and prescrib…
- 0 votes0 replies0 views
Max-function formulation of the ReLU depth-separation conjecture
Let , and let be the set of continuous piecewise linear functions on computable with hidden layers. Max-function formulation of…
- 0 votes0 replies0 views
Hertrich–Lindner–Raginsky–Bach depth-separation conjecture for ReLU networks
Let denote the set of continuous piecewise linear functions defined on and computable with hidden layers. Hertrich–Lindner–Raginsky–Ba…
- 0 votes0 replies0 views
The chaotic-behavior conjecture for neural-network depth-width limitations
Consider neural networks representing functions, with depth measuring the number of compositional layers and width measuring the number of units available in a layer. The source di…
- 0 votes0 replies0 views
The chaotic-complexity conjecture for depth-width trade-offs in neural networks
A neural network's chaotic behavior refers to the dynamical complexity of the maps arising in its depth-composition structure, while the notions of complexity used in depth-separat…